Captioning podcast clips with an API: settings that read well

A podcast clip has little to look at besides the speaker or a waveform, so the captions do most of the visual work. Most people watch these with the sound off. The goal is text that is easy to read at a glance and does not fight the speaker's face.

The request

{
  "source_url": "https://example.com/podcast-clip.mp4",
  "options": {
    "font": "inter",
    "size": "medium",
    "max_words": 3,
    "max_lines": 2,
    "position": "bottom"
  }
}

Why these settings

• max_words 3 gives a calm pace that follows conversation. One word at a time feels frantic for dialogue, and long groups are hard to read as they change.
• size medium keeps the text from covering faces on a talking head. Go large on vertical clips that will be watched on a phone.
• position bottom is the natural spot for speech, but on vertical video try a fraction near 0.7 so platform buttons do not cover the text.
• font: pick a plain, readable typeface for long speech. Save the heavy display fonts for short punchy clips. See the docs for the full list of 18 fonts.

Two things to check

Pick a font that looks clear over both light and dark footage, since podcast clips often cut between cameras. And run one real clip through before you batch a whole episode, so the look is settled once.

caption.sh burns the captions into the video, word by word, so the file looks the same wherever you upload it. It does not produce a separate caption file.